Research On DDoS Attack Detection Technology Based On Improved CNN

The security issues caused by distributed denial of service(DDoS)attacks are endless,and how to quickly and effectively detect DDoS intrusion attacks is a serious problem facing the current field of network security.Deep learning technology is widely used in research on DDoS attack detection due to its powerful data processing ability and adaptability.However,the commonly used deep learning models have many limitations in the application of resource constrained intelligent terminal devices.In order to further improve the application effect of deep learning technology in the field of DDoS attack detection.This article conducts research from three aspects: data processing,lightweight,and model improvement.A DDoS attack detection method based on Lightweight Convolutional Neural Networks(LCNN)and a DDoS attack detection method based on self attention mechanism and LCNN have been proposed.Relevant experiments have demonstrated the effectiveness of the proposed method.The main work of this article is as follows:(1)Addressing the problems of low accuracy and slow speed in traditional DDoS attack detection methods.This article proposes a DDoS attack detection method based on lightweight convolutional neural networks.In the data processing stage,this method aims to address the problem of low normal traffic in the CICDDoS2019 dataset.It first captures a portion of normal access traffic and adds it to the dataset.Then,based on the importance of relevant targets,the data content is identified,and the corresponding data in the dataset is extracted.Then,packet attribute statistics are used to enable the convolutional layer to learn the correlation between data packets within the same class,thereby improving classification performance.In order to improve the robustness and processing speed of the model,a lightweight convolutional neural network(LCNN)with two consecutive large kernel convolutional layers was adopted.Finally,it was validated on the dataset.The experimental results showed that the method proposed in this paper is more effective and accurate in detecting DDoS attacks.(2)This paper proposes a DDoS attack detection method based on self attention mechanism and LCNN to address the requirements of low latency,low computational complexity,and high detection rate in the research of intrusion detection systems(IDS)using lightweight deep learning methods.This method considers data processing,model quantization,and lightweight aspects.In order to further improve model efficiency,detection accuracy,and enhance the application ability of deep learning in resource constrained devices and environments,attention mechanism is introduced.Finally,deploy and test the trained model on Raspberry Pi.The experimental results show that the method proposed in this article is more effective and accurate in DDoS intrusion detection,with an average detection rate of 98%.The DDoS attack detection method based on deep learning proposed in this article can effectively address the issues of low intrusion detection accuracy and slow detection speed,and there is room for application in the security field.